HSC preparation in Bangladesh is a content-availability problem that has been solved many times — and a continuity problem that nobody had solved. Students collect questions from photocopies, run drills in standalone MCQ apps, watch explanation videos on YouTube, and track their progress in a paper notebook. Each tool is decent at one task; none of them know what the others know about the student.

Prottoy's bet was simple: hold the curriculum, the attempts, and the recommendations in one place — on the device the student actually uses (a mid-range Android phone) and the network they actually have (variable mobile data on a metered plan). Practice volume — and the consistency of it — compounds from there.

Flutter for the mobile app gave us iOS and Android from one codebase without the rendering compromises common to cross-platform alternatives on mid-range Android. The animation budget on three-year-old phones is where most cross-platform apps lose their feel; Flutter's compiled UI kept the practice flow at 60fps on devices that age.

Supabase carries auth, the question bank, attempt history, and analytics from day one — one backend across mobile and web, no separate infrastructure to run. The Next.js marketing site at prottoy.app is static-first for SEO, with a dashboard handoff that uses the same Supabase tokens as the mobile app. One identity, three surfaces.

Offline-first is the constraint everything else bends to. Questions and attempts cache locally; the app works on a flaky connection and syncs when the network returns. A typical study session uses under 5 MB of data — meaningful on a 1 GB monthly plan.

The first build let students pick a subject and get random questions. Beta users wanted more direction. Version two introduced an adaptive mode that surfaces topics where the student's accuracy is lowest, weighted by exam priority. Practice volume per student climbed sharply in the first month it shipped.

Auth went through three iterations. Email asked for something nobody used. SMS one-time codes worked in Dhaka but failed in remote districts because of carrier delays — sign-up abandonment crossed 30%. The current flow is phone-number-plus-passcode set on the device, with no SMS dependency. Sign-up completion is now above 90%.

HSC is where we are today. The architecture — content as data, attempts as events, adaptation as a function of the two — generalises cleanly. SSC, JSC, and university admission test prep are the next three rooms to unlock.